Chapter 16 Quiz: Shot Quality Models

Instructions

Answer all 25 questions. Each question is worth 4 points (100 total).

Section A: Conceptual Understanding (Questions 1-10)

Question 1

What is the fundamental principle behind expected points (xPoints)?

A) Points scored divided by games played B) Probability of making a shot multiplied by point value C) Total points divided by total shots D) The average points per possession

Question 2

Why do corner three-pointers typically have higher expected value than above-the-break threes?

A) They are worth more points B) The angle makes them easier to shoot C) They have a shorter distance to the basket D) Defenders are slower to close out

Question 3

In shot quality models, what does "defender distance" measure?

A) Distance between defenders B) Distance from shooter to nearest defender at release C) Distance the defender traveled D) Distance from defender to basket

Question 4

What is the primary advantage of using logistic regression for shot prediction?

A) It runs faster than other models B) It outputs probabilities between 0 and 1 C) It doesn't require training data D) It handles missing data automatically

Question 5

What does a positive "shot-making above expected" indicate?

A) Player takes easy shots B) Player takes difficult shots C) Player makes shots at a higher rate than the model predicts D) Player takes more shots than teammates

Question 6

Why is model calibration important for shot quality models?

A) It makes the model run faster B) It ensures predicted probabilities match actual outcomes C) It reduces the amount of data needed D) It eliminates the need for cross-validation

Question 7

What is the relationship between touch time and shooting efficiency?

A) Longer touch time improves accuracy B) Touch time has no effect on accuracy C) Shorter touch time (catch-and-shoot) tends to have higher FG% D) Only affects three-point shooting

Question 8

What does the Brier score measure in shot quality models?

A) Model speed B) Feature importance C) Mean squared error of probability predictions D) Number of correct predictions

Question 9

Why might random train-test splits be problematic for basketball shot data?

A) They are too slow to compute B) They may leak information across related shots C) They require too much memory D) They don't work with categorical variables

Question 10

What is "shot quality differential"?

A) Difference between two players' shot totals B) Difference between actual and expected shooting percentage C) Difference in shot selection between quarters D) Difference between home and away shooting

Section B: Calculations (Questions 11-18)

Question 11

A player takes a shot with 42% probability from two-point range. What is the expected points?

A) 0.42 points B) 0.84 points C) 1.26 points D) 2.00 points

Question 12

Given: Corner 3 (39% FG) vs Mid-range (42% FG). Which has higher expected value?

A) Corner 3 (1.17 xPts) B) Mid-range (0.84 xPts) C) They are equal D) Cannot determine without more information

Question 13

A model predicts P(make) = 0.60 and the shot is missed. What is the log loss contribution?

A) -ln(0.60) B) -ln(0.40) C) 0.60 D) 0.40

Question 14

Player shoots 45% on 200 attempts. Model expected 42% FG. How many makes above expected?

A) 3 makes B) 6 makes C) 9 makes D) 12 makes

Question 15

Logistic model: P(make) = 1/(1 + exp(-(0.5 - 0.05*dist))). What is P(make) at 10 feet?

A) 50% B) 55% C) 60% D) 65%

Question 16

Team takes 30 shots at rim (65% FG), 40 mid-range (40% FG), 30 threes (36% FG). Expected total points?

A) 95.3 points B) 103.4 points C) 111.6 points D) 119.2 points

Question 17

A player's xPts/shot is 1.05, volume is 12 FGA. Another player has 1.10 xPts/shot, 8 FGA. Who creates more expected points per game?

A) Player 1 (12.60 xPts) B) Player 2 (8.80 xPts) C) They are equal D) Cannot determine

Question 18

Model predicts 100 shots at 50% probability, 48 go in. What is the calibration error for this bucket?

A) -2% B) +2% C) -4% D) +4%

Section C: Model Design (Questions 19-22)

Question 19

Which feature would likely be MOST important in a shot quality model?

A) Player jersey number B) Distance to basket C) Game day of week D) Arena temperature

Question 20

What is the purpose of cross-validation in shot quality modeling?

A) To speed up training B) To estimate model performance on unseen data C) To reduce data storage needs D) To eliminate outliers

Question 21

When should you include "shooter ability" as a feature in a shot quality model?

A) Always - it improves predictions B) Never - it's not measurable C) Depends on the use case (prediction vs. shot quality evaluation) D) Only for three-point shots

Question 22

What problem does ridge regression solve in shot quality models with many features?

A) Missing data B) Overfitting C) Slow computation D) Class imbalance

Section C: Applications (Questions 23-25)

Question 23

A defense reduces opponent xPts from 1.05 to 0.95. Over 100 possessions, how many points saved?

A) 5 points B) 10 points C) 15 points D) 20 points

Question 24

Player A has xFG% of 48% and actual FG% of 46%. Player B has xFG% of 44% and actual FG% of 45%. Who is the better shot-maker relative to difficulty?

A) Player A B) Player B C) They are equal D) Cannot determine

Question 25

A team's lineup generates 1.12 xPts but scores only 1.06 actual. What does this suggest?

A) Good shot selection, poor execution B) Poor shot selection, good execution C) Good at both D) Poor at both

Answer Key

Scoring Guide

• 90-100: Excellent understanding of shot quality concepts
• 80-89: Good grasp with minor gaps
• 70-79: Adequate understanding, review weak areas
• 60-69: Needs significant review
• Below 60: Re-read chapter before proceeding